Currently, the proportion of invariant sites is not considered while computing
edge likelihoods.
We can instead create an extra "dummy" rate category with weight p (whereas the
original N category weights are now (1-p)/N). In other words, likelihoods for
the dummy category are still computed for all sites.
Original issue reported on code.google.com by imran.fa...@gmail.com on 18 Sep 2013 at 3:22
Original issue reported on code.google.com by
imran.fa...@gmail.com
on 18 Sep 2013 at 3:22